Profiles that update as you talk
An AI-native talent network where candidates and roles are living, evidence-linked profiles, scored in both directions, and introduced only when both sides say yes.
Software & AI careers
Kif helps software and AI professionals in Pakistan and remote roles worldwide, and the teams hiring them, compare skills, experience, and preferences. Learn how matching works, or explore the other channels available for your search.
Skills include Python, FastAPI, Kubernetes, React, TypeScript, PostgreSQL, Kafka, PyTorch, Terraform, Figma, Product discovery, Flutter, Playwright, Solidity, Rust, Go, Spark, dbt, Accessibility, Incident management, Unity, Technical writing. Roles include Senior Backend Engineer, Data Engineer, ML Engineer, Site Reliability Engineer, Product Designer, Frontend Engineer, Engineering Manager, QA Automation Engineer, Mobile Engineer, Customer Success Engineer, Product Manager, Security Engineer.
The problem
Both sides lose. Candidates are reduced to a document and judged by word overlap. Hiring teams drown in applications they can't evaluate and reject people they never understood.
It can't carry evidence, preferences or constraints, and nobody updates it for a new project, a move, or a change of mind.
Word overlap is not competence. Strong people vanish; whoever stuffed the right words surfaces. Run the simulation and watch.
Candidates get no explanation and no feedback. Recruiters get volume they can't read and decisions they can't defend.
Scanning for the words in the job post…
strong candidate kept, rightly kept on words alone✕ rejected
The shift
Drag the handle. Same candidate, same role, two worlds.
Ayesha_CV_final_v3.pdf · 14 months old
JD_Senior_Backend.docx
Boolean search
python AND django AND "5+ years"
0 of 1,284 shortlisted
Applicant view
Application received. Status: no response · day 23
Auto-reply · do not reply
Unfortunately we will not be moving forward with your application at this time. (Reason: none given.)
Recruiter inbox · this week
Senior backend engineer
role · Northwind
Senior Backend Engineer, Payments
role
you
Why · shown to both sides
How it works
I lead a platform team of 6. Last year we cut deploy time from 40 to 8 minutes.
Nice, that's a concrete result. What did you personally own in that change, and which tools did the team use?
One question at a time. Facts and aspirations kept apart.
01 · Conversation
No forms. You describe your work, or the role you need to fill, the way you'd explain it to a colleague. The assistant asks one good question at a time, and asks before saving anything consequential.
02 · Living twin
Every fact is structured, linked to the evidence behind it, and versioned. Say "I moved to Islamabad" six months later and only that field changes. The transcript is never the profile.
03 · Two-way match
Hard requirements are checked first. Then a deterministic engine scores fit in both directions and shows its reasons: strong evidence, gaps, and what is still unknown. A one-sided match can't hide behind an average.
04 · Mutual introduction
Either side can express interest. When it's mutual, each person confirms before any contact details are shared. From there, every decision is made by a human.
The science of the match
Decades of selection research point the same way: structured, behaviour-based evidence predicts how someone will do far better than unstructured impressions or keyword overlap. The engine is built on that, and it shows its work.
Every role competency gets observable 1–5 anchors ("designs idempotent services for one domain without guidance"), the way structured assessment has done it for decades. Adjectives like "rockstar" are not a requirement.
The assistant probes the way a good structured interview does: situation, what you personally did, the measurable result. Claims are labelled verified, corroborated, self-reported or unknown, and never dressed up.
Person–job fit runs both ways. Hire fit asks whether you can do this role; job fit asks whether this role is right for you. They are combined with a harmonic mean, so a one-sided match can't hide behind an average.
Score anatomy
scoring.v1Hire fit · can they do this role?weights sum to 100
scored against the role's 1–5 behavioural anchors, weighted by the recruiter
exact skills first, adjacent skills (FastAPI ≈ Django) at partial credit
specific, checkable claims outrank vague self-ratings
has the person done this kind of work before
the team's stated preferences, never hard filters
Job fit · is this role right for them?weights sum to 100
where you said you want to go, and what you want to avoid
does the range reach your floor and your target
remote, hybrid or on-site, as you prefer
a step up, the same level, or a step down
employment type, hours, and the rest
mutual = harmonic_mean(hire_fit, job_fit) × confidence
A harmonic mean punishes lopsided matches: 90 and 30 give 45, not 60. Confidence drops when facts are unknown or evidence is thin, so a high score always means the engine knew enough to say so.
kif.app/p/8k2…
A random token, never your ID or email
Has Ayesha built payment systems?
Yes. Self-reported with specific evidence: she cut p95 payment API latency from 900 ms to 220 ms at PayFlow (2021–present).
I'm an AI representation based on what Ayesha allowed this page to share.
Your profile, with its own AI representative
Every profile and every published role can have a short public link and a QR code. Behind it is a read-only AI that knows only what you chose to share, and says so.
Discloses that it's an AI
First message, every time. It never pretends to be you.
Answers only from what you allowed
Private fields, pay, contact details and other people's data are out of reach, even under prompt injection.
Known, derived, or unknown
It labels what it knows, what it inferred, and what you didn't share.
Revoke in one click
Rotate or kill the link and the page goes dark. Rate limits and a daily budget stop scraping.
0
fit scores per match
one for the role, one for you
0
profile that grows with you
import, review, and refine
0%
of decisions made by people
the engine ranks, it never rejects
0
confirmations before contact
one from each side
Two sides, one network
For candidates
For hiring teams
Principles that don't drift
Each one is enforced in code, not promised in a prompt.
same profile v13 + same role v4
Filters, scores and ranking are reproducible code, not a model's mood. The same profile and role always get the same result.
role
you
average 60
harmonic 45
lopsided stays visible
Fit for the role and fit for the candidate are scored separately and combined with a harmonic mean, so a one-sided match can't hide.
"Cut p95 latency from 900 ms to 220 ms"
Every material claim is labelled, and never dressed up as more than it is.
ranked
Candidate 3F9A2C · no name, no pay
Pseudonymous until mutual interest; names and contact details only after both confirm.
The system ranks, explains and introduces. It never rejects anyone, and protected traits are never features.
I moved to Islamabad.
v12 · location: Lahore
Say one sentence and only the affected fields change. Matches refresh in the background within minutes.
Two minutes to a living profile or a structured role. No forms, no black boxes, no surprises about who sees what.